Dermatologist-level classification of skin cancer with deep neural networks

Over the past few decades, cases of skin cancer have increased enormously. It is getting most popular cause of deaths in men/woman's even in both developed and undeveloped countries. If skin cancer can't be treated in early stages, it can go down in human skin and spread quickly to other parts in the body. Melanoma causes luckily become visible to sufferers creating an opportunity to detect in early stages. The recent advent of deep learning methods for computer - aided diagnosis has allowed the development of intelligent healthcare imaging-based diagnostic systems that can facilitate the human experts in making informed decisions about a patient's condition. In this report will focus on the problem of skin lesion classification, specifically the detection of skin cancer (melanoma) in early stages, and present a deep learning (Convolution Neural Network) based approach to classify the dermoscopic image containing a skin lesion as cancerous or benign. We have identified what methodology and parameters to use to replicate this study. In this report we have tried to brought in light that how deep networks and convolutional neural networks are taking the place of handcrafted feature extractors in different image classifications.